採用標記後重複觀察之資料以估計黑面琵鷺之存活率 暨推估族群數量

Abstract

生物族群的動態可以利用該物種之生活史參數來研究。 事實上, 因為動物的標記試驗無法長時間密集地監控, 故個體的正確死亡時間和地點通常無法得知。 因此, 可透過捕捉再看的統計模式來估計動物的存活率和有關的參數。 本論文提供詳盡的模式方法介紹, 且實際分析野外收集之數據並作解釋。 文章中分析黑面琵鷺 11 年的標記再看資料, 此鳥種的背景知識在內文中的第二章有概要性的介紹。 在第三章中, 我們先將完整資料分成三個群別(幼鳥/健康, 成鳥/健康, 成鳥/復原), 之後採用開放族群之 Cormack-Jolly-Seber 模式進行分析。 此方法是從個體在第一次被捕捉的條件下, 估計其物種的存活率和再看機率, 文中的計算步驟均以 MARK 分析軟體執行。 最後, 我們解釋;討論三群別的參數估計值。 在第四章中, 我們詳細地描述如何改善模式中過多參數的問題。 利用 likelihood-ratio test 及 Akaike's Information Criterion 來選擇最佳的簡約模式。 在黑面琵鷺的例子中, 最佳的模式僅可由一個存活率, 0.8568(95%CI: 0.7287-0.9302), 和一個再看機率, 0.9333(0.7851-0.9817),來描述。 接下來, 我們在模式中加入限制條件的方法, 以估算參數受生理或環境因素的影響(年齡或感染肉毒桿菌)。 當個體遭受疾病感染時, 其存活率約下降百分之40。 此外, 個體的再看機率在幼鳥時期約比成鳥時期減少百分之10。 在第五章中, 我們利用前幾章所估出的存活率來建立黑面琵鷺在台灣的族群數量模式, 並預測其未來的族群趨勢。 在本章中, 決定性和隨機性模式都放入討論, 且進一步作模擬分析。 由模式分析結果, 當時間來到 2015年時, 黑面琵鷺在台灣的數量(+-SD)將成長到 2360 +- 275 (今日其在台灣的數量為1104)。 在維持現今的條件下, 黑面琵鷺將不會有滅亡的危機。 但當其數量成長兩倍時, 活動空間和食物將是需要關心的議題。

The understanding of the dynamics of animal populations and of related ecological and evolutionary issue frequently depends on analyses of individuals' life history parameters. Because marked individuals cannot be followed closely through time, the exact time of death is most often unknown. Thus, the analysis of survival studies and experiments must be based on capture-recapture(or resighting) models. This article presents a detailed, practical example on the design, analysis, and interpretation of capture-recapture studies. The marked-resighting data set on the black-faced spoonbill is given to illustrate the theory, and its lifestyle of backgrounds is covered in detail in Chapter 2. In Chapter 3 we consider time-dependent Cormack-Jolly-Seber open population models with groups of animals, which are central to the article. This approach is conditioning on first capture; hence it dose not attempt to model the initial capture of unmarked animals as functions of population abundance in addition to survival and resighting probabilities which were developed and estimated using MARK. The fluctuations of estimates in three groups of birds (juvenile/health, adult/health and adult/recovery) are compared. In Chapter 4 we give a detailed description and demonstration of model selection. Goodness of fit, likelihood-ratio test and Akaike's Information Criterion are introduced for the selection of more parsimonious models. The best model to describe the data of BFS is one single survival rate, 0.8568(95%CI: 0.7287-0.9302), and one single resighting probability, 0.9333(0.7851-0.9817). Next, we examine the effects of physical situation or environmental event(outbreak of botulism or age) by adding constraints in the models. Suffering from diseases the survival rate of BFS drop about 40 percent. Resighting probabilities in the earlier three years are 10 percent lower than latter years. In Chapter 5 we apply annual population information and life history parameters estimated in previous chapters to model and predict the population of BFS, and evaluate the time of extinction. Deterministic and stochastic models are both considered, and simulated results are provided. When the time is in the year of 2015, the number(+- SD) of BFS in Taiwan will grow up to 2360 +- 275 (today's population in Taiwan is 1104), and in current situation the possibility to be extinct is quite small. However, when the number is double, space and food will be another big issue to be considered.